Stroke risk intelligent stratification and early warning system based on carotid plaque characteristics

Abstract

The present application relates to the technical field of image analysis, in particular to a stroke risk intelligent stratification early warning system based on carotid plaque characteristics, which can realize the following steps through the mutual cooperation between multiple modules: obtaining the target SWI image sequence, the target T1 image sequence and the target T2 image sequence after multi-sequence registration corresponding to the patient to be detected, and performing region extraction; analyzing according to the gray value and the gradient value to determine the initial calcified plaque characteristic factor; performing calcification stability analysis according to the initial calcified plaque characteristic factor to obtain the target calcified plaque characteristic factor; thereby determining the vulnerable hemorrhagic plaque possibility corresponding to each initial region of the carotid plaque, which is used for assisting in the stratification early warning of stroke risk. The present application quantifies the vulnerable hemorrhagic plaque possibility, which can accurately assist in the identification of calcified plaque and vulnerable hemorrhagic plaque to a certain extent, improves the identification accuracy, and thus improves the auxiliary effect of stroke risk early warning.

Description

Technical Field

[0001] This invention relates to the field of image analysis technology, and more specifically to an intelligent stratified early warning system for stroke risk based on carotid plaque characteristics. Background Technology

[0002] With the development of technology, image analysis is being applied more and more widely. For example, it can be used to identify carotid artery plaques to assist in stroke risk stratification and early warning. Since calcified carotid artery plaques and hemorrhagic carotid artery plaques correspond to different stroke risks, it is often necessary to distinguish the state of the carotid artery plaque—whether it is calcified or hemorrhagic—when assisting in stroke risk stratification and early warning. Currently, the common method for object recognition is to identify different objects based on their grayscale values.

[0003] However, when identifying calcified carotid plaques and hemorrhagic carotid plaques based on different gray values, the following technical problems often arise: Since the grayscale difference between calcified carotid plaques and hemorrhagic carotid plaques is often not obvious in images, considering only the difference in grayscale values ​​when distinguishing between calcified and hemorrhagic carotid plaques may lead to misjudgment, resulting in poor accuracy in identifying different carotid plaques and consequently poor auxiliary effect in stroke risk warning. Summary of the Invention

[0004] To address the technical problem of poor stroke risk warning due to the low accuracy of different carotid plaque identification, this invention proposes an intelligent hierarchical early warning system for stroke risk based on carotid plaque characteristics.

[0005] This invention provides an intelligent stratified early warning system for stroke risk based on carotid plaque characteristics. The system includes: The image acquisition and region extraction module is used to acquire the target SWI image sequence, target T1 image sequence and target T2 image sequence corresponding to the patient to be tested after multi-sequence registration, and to extract the region of each image in these image sequences to obtain the initial region of carotid plaque; The grayscale and gradient analysis module is used to analyze the grayscale and gradient values ​​of the pixels in the initial region of each carotid plaque to determine the initial calcified plaque feature factors corresponding to the initial region of each carotid plaque. The calcification stability analysis module is used to perform calcification stability analysis based on the initial calcification plaque characteristic factors corresponding to different initial regions of carotid plaques, and to obtain the target calcification plaque characteristic factors corresponding to each initial region of carotid plaque. The vulnerable hemorrhage plaque probability determination module is used to determine the probability of vulnerable hemorrhage plaques corresponding to each initial region of carotid artery plaque based on the distribution differences of the initial regions of carotid artery plaques in the target SWI image sequence, target T1 image sequence, and target T2 image sequence, as well as the target calcified plaque feature factors corresponding to each initial region of carotid artery plaque, in order to assist in stroke risk stratification and early warning.

[0006] Optionally, the step of performing region extraction on each image in these image sequences to obtain the initial region of the carotid plaque includes: Threshold segmentation is performed on each image in the target SWI image sequence, target T1 image sequence, and target T2 image sequence to obtain the initial region of the carotid plaque that characterizes the carotid plaque.

[0007] Optionally, the step of analyzing the grayscale values ​​and gradient values ​​corresponding to pixels within the initial region of each carotid plaque to determine the initial calcified plaque feature factors corresponding to each initial region of the carotid plaque includes: An initial region of any carotid artery plaque is defined as a labeled region, and the gray value and gradient value corresponding to each pixel in the labeled region are normalized to obtain the gray value normalization factor and gradient normalization factor corresponding to each pixel in the labeled region. The initial calcification patch feature factor corresponding to the marked region is determined based on the mean and standard deviation of the gray-level normalization factor corresponding to all pixels in the marked region, and the mean of the gradient normalization factor corresponding to all edge pixels in the marked region.

[0008] Optionally, the step of performing calcification stability analysis based on the initial calcified plaque characteristic factors corresponding to different initial regions of carotid plaques to obtain the target calcified plaque characteristic factors corresponding to each initial region of carotid plaque includes: Any initial region of a carotid artery plaque is identified as a labeled region, and the image sequence to which the labeled region belongs is identified as a labeled image sequence. Each image sequence in the target SWI image sequence, target T1 image sequence, and target T2 image sequence, except for the labeled image sequence, is identified as a reference image sequence, resulting in two reference image sequences. From each reference image sequence, images with the same image number as the image to which the marked region belongs are selected as reference images, resulting in two reference images; Initial carotid plaque regions that match the marked regions are selected from each reference image to obtain the matching regions corresponding to each reference image; The matching regions corresponding to all reference images are used to form a set of matching regions corresponding to the marked region; If the number of matching regions in the matching region set is less than 2, then the target calcified plaque feature factor corresponding to the marked region is set to the preset calcified plaque feature factor. If the number of matching regions in the matching region set is equal to 2, then the target calcified plaque feature factor corresponding to the marked region is determined based on the mean and range of the initial calcified plaque feature factors corresponding to the marked region and the two matching regions in the matching region set.

[0009] Optionally, the step of selecting initial carotid plaque regions that match the marked region from each reference image to obtain a matching region corresponding to each reference image includes: Any reference image is selected as the marker reference image, and the region with the same position as the marker region is selected from the marker reference image as the marker representative region; Initial regions of carotid plaques that do not intersect with the marked regions in the labeled reference image are selected as candidate regions. If no candidate region exists, it is determined that the labeled reference image does not have a corresponding matching region; If a candidate region exists, the candidate region with the largest intersection area with the region represented by the marker is selected from all candidate regions and used as the matching region corresponding to the marker reference image.

[0010] Optionally, determining the likelihood of a vulnerable hemorrhagic plaque corresponding to each initial carotid plaque region based on the distribution differences of the initial carotid plaque regions in the target SWI image sequence, target T1 image sequence, and target T2 image sequence, and the target calcified plaque characteristic factors corresponding to each initial carotid plaque region, includes: An initial region of any carotid artery plaque is identified as a marker region, and a set of reference regions corresponding to the marker region is selected from the image sequence to which the marker region belongs. If the set of reference regions corresponding to the marked region is empty, then a set containing only a preset continuity index is constructed as the set of vertical spatial continuity indices corresponding to the marked region. If the set of reference regions corresponding to the marked region is not empty, then the longitudinal spatial continuity index corresponding to each reference region is determined according to the area overlap between the marked region and each reference region in the set of reference regions, and the set of longitudinal spatial continuity indices corresponding to the marked region is obtained. Based on the differences in the distribution of the initial region of carotid plaque in the target SWI image sequence, target T1 image sequence, and target T2 image sequence, and the set of longitudinal spatial continuity indices corresponding to the marked region, the initial vulnerable bleeding factor corresponding to the marked region is determined. The likelihood of a vulnerable hemorrhage plaque corresponding to the marked region is determined based on the initial vulnerable hemorrhage factor and the target calcified plaque characteristic factor corresponding to the marked region.

[0011] Optionally, the step of filtering the reference region set corresponding to the marked region from the image sequence to which the marked region belongs includes: The image sequence to which the marked region belongs is determined as the marked image sequence, and the image to which the marked region belongs is determined as the undetermined image; The image preceding and following the image to be determined in the marked image sequence is identified as a temporary image, and the region with the same position as the marked region is selected from the temporary image as a temporary representative region. Initial carotid plaque regions with non-empty intersections with the temporary representative region are selected from the temporary image and used as reference regions; The reference region with the largest intersection area with the temporary representative region is selected from all reference regions and used as the reference region corresponding to the temporary image, thereby obtaining the reference regions corresponding to the previous image and the next image of the image to be determined. The reference regions corresponding to the preceding and following images of the image to be determined are used to form a set of reference regions corresponding to the marked region.

[0012] Optionally, determining the vertical spatial continuity index corresponding to each reference region based on the area overlap between the marked region and each reference region in the corresponding set of reference regions includes: The area of ​​the intersection between the temporary representative area and each reference area is determined as the target overlap area between the marked area and each reference area; Based on the target overlap area between the marked region and each reference region, and the difference between the initial calcified plaque characteristic factors corresponding to the marked region and each reference region, the longitudinal spatial continuity index corresponding to each reference region is determined.

[0013] Optionally, determining the initial vulnerable bleeding factor corresponding to the marked region based on the distribution differences of the initial carotid plaque region in the target SWI image sequence, target T1 image sequence, and target T2 image sequence, and the set of longitudinal spatial continuity indices corresponding to the marked region, includes: From all initial regions of carotid plaques in the target SWI image sequence, target T1 image sequence, and target T2 image sequence, the initial regions of carotid plaques that characterize the carotid plaques to which the labeled region belongs are selected to form the labeled SWI region set, labeled T1 region set, and labeled T2 region set corresponding to the labeled region; Regions with the same image sequence number in the labeled SWI region set, labeled T1 region set, and labeled T2 region set are formed into target region groups. The number of elements in different target region groups represents the differences in the distribution of the initial region of carotid plaque in different image sequences. Based on the area of ​​each target region group and its corresponding set of longitudinal spatial continuity indices, the local vulnerable bleeding factors corresponding to each target region group are determined. Based on the mean value of the local vulnerable bleeding factors corresponding to all target region groups and the number of target region groups with an element count of 3, the initial vulnerable bleeding factor corresponding to the carotid plaque to which the marked region belongs is determined. The initial vulnerable bleeding factor corresponding to the carotid artery plaque to which the marked region belongs is denoted as the initial vulnerable bleeding factor corresponding to the marked region.

[0014] Optionally, determining the local vulnerable bleeding factor corresponding to each target region group based on the area of ​​the region within each target region group and its corresponding set of longitudinal spatial continuity indices includes: Any target region group is designated as a temporary region group. If there are fewer than 3 elements in the temporary region group, the local vulnerable bleeding factor corresponding to the temporary region group is set as a preset vulnerable bleeding factor. If there are 3 elements in the temporary region group, the T1 continuity difference corresponding to the temporary region group is determined based on the difference between the set of longitudinal spatial continuity indices corresponding to the SWI region and the set of longitudinal spatial continuity indices corresponding to the T1 region; the T2 continuity difference corresponding to the temporary region group is determined based on the difference between the set of longitudinal spatial continuity indices corresponding to the SWI region and the set of longitudinal spatial continuity indices corresponding to the T2 region; and the local vulnerable bleeding factor corresponding to the temporary region group is determined based on the difference between the area of ​​the SWI region and the areas of the T1 and T2 regions, as well as the T1 and T2 continuity differences corresponding to the temporary region group.

[0015] The present invention has the following beneficial effects: The intelligent stratified early warning system for stroke risk based on carotid plaque characteristics of the present invention quantifies the probability of vulnerable bleeding plaques and can accurately assist in the identification of calcified plaques and vulnerable bleeding plaques to a certain extent. It solves the technical problem that the auxiliary effect of stroke risk warning is poor due to the poor accuracy of different carotid plaque identification, improves the accuracy of identification, and thus improves the auxiliary effect of stroke risk warning. Specifically, this invention quantifies the initial calcified plaque characteristic factors corresponding to the initial region of the carotid plaque by analyzing the grayscale and gradient of the initial region of the carotid plaque, and performs calcification stability analysis to quantify the target calcified plaque characteristic factors corresponding to the initial region of the carotid plaque. Based on the distribution differences of the initial region of the carotid plaque in the target SWI image sequence, target T1 image sequence, and target T2 image sequence, and the target calcified plaque characteristic factors corresponding to the initial region of the carotid plaque, the probability of the vulnerable hemorrhagic plaque corresponding to the initial region of the carotid plaque is quantified. The larger the value, the more likely the initial region of the carotid plaque is to represent a vulnerable hemorrhagic plaque. To a certain extent, this can accurately assist in the identification of calcified plaques and vulnerable hemorrhagic plaques, improve the accuracy of identification, and thus improve the auxiliary effect of stroke risk warning. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the structure of an intelligent stratified early warning system for stroke risk based on carotid plaque characteristics according to the present invention; Figure 2 This is a flowchart of an intelligent stratified early warning method for stroke risk based on carotid plaque characteristics according to the present invention; Figure 3 This is a schematic diagram of the structure of a computer device according to the present invention. Detailed Implementation

[0017] refer to Figure 1 This diagram illustrates the structure of an intelligent stratified early warning system for stroke risk based on carotid plaque characteristics, according to the present invention. The intelligent stratified early warning system for stroke risk based on carotid plaque characteristics includes: The image acquisition and region extraction module 101 is used to acquire the target SWI image sequence, target T1 image sequence and target T2 image sequence corresponding to the patient to be tested after multi-sequence registration, and to extract the region of each image in these image sequences to obtain the initial region of the carotid plaque.

[0018] The patient to be tested can be one undergoing stroke risk stratification and early warning. Multi-sequence registration enables the registration of multiple image sequences, allowing images representing the same anatomical location of the patient in different image sequences to correspond spatially. It maps different image sequences to the same three-dimensional physical coordinate system, and through resampling, ensures that voxels in the same spatial coordinates can represent the same anatomical location of the patient, ultimately enabling images with the same sequence number to represent the same location. For example, multi-sequence registration can be, but is not limited to: rigid registration based on mutual information, affine registration based on normalized mutual information, and non-rigid registration based on B-spline. Images with the same sequence number in the target SWI image sequence, target T1 image sequence, and target T2 image sequence can represent the same location of the patient. Images in the target SWI image sequence, target T1 image sequence, and target T2 image sequence are also called cross-sectional images or slice images. The initial region of the carotid plaque can be the slice region containing the carotid plaque to be classified as either a calcified plaque or a hemorrhagic vulnerable plaque.

[0019] As an example, the image acquisition and region extraction module 101 specifically implements the following steps: The first step is to obtain the target SWI image sequence, target T1 image sequence, and target T2 image sequence of the patient to be tested after multi-sequence registration.

[0020] For example, an MRI (Magnetic Resonance Imaging) scanner can be used to scan and obtain SWI (Susceptibility Weighted Imaging) image sequences, T1 (T1-weighted Imaging) image sequences, and T2 (T2-weighted Imaging) image sequences of the patient to be examined. Based on mutual information, rigid registration is performed to map the SWI, T1, and T2 image sequences to the same three-dimensional physical coordinate system. Resampling is then used to ensure that the voxels of the different image sequences obtained in the same spatial coordinate system can represent the same anatomical location of the patient. The different image sequences obtained in the end can be referred to as the target SWI image sequence, the target T1 image sequence, and the target T2 image sequence, respectively.

[0021] The second step is to extract regions from each image in these image sequences to obtain the initial regions of carotid plaques.

[0022] For example, threshold segmentation can be performed on each image in the target SWI image sequence, target T1 image sequence, and target T2 image sequence to identify the region representing the carotid plaque, which is denoted as the initial region of the carotid plaque, thus obtaining the initial region of the carotid plaque representing the carotid plaque.

[0023] It should be noted that identifying regions characterizing carotid artery plaques is a relatively common area identification task. Therefore, existing algorithms capable of region identification can be used to perform carotid artery plaque identification. Besides threshold segmentation, existing algorithms for region identification include watershed algorithms, clustering algorithms, region growing algorithms, and neural network techniques. Therefore, identifying regions characterizing carotid artery plaques can be achieved not only through threshold segmentation but also through watershed algorithms, clustering algorithms, region growing algorithms, or neural network techniques.

[0024] The grayscale and gradient analysis module 102 is used to analyze the grayscale and gradient values ​​of the pixels in the initial region of each carotid plaque to determine the initial calcified plaque feature factors corresponding to the initial region of each carotid plaque.

[0025] As an example, determining the initial calcified plaque characteristic factors corresponding to the initial region of each carotid plaque may include the following steps: The first step is to determine any initial region of a carotid artery plaque as the marked region, and then normalize the gray value and gradient value corresponding to each pixel in the marked region to obtain the gray normalization factor and gradient normalization factor corresponding to each pixel in the marked region.

[0026] For example, the formula for determining the gray-level normalization factor corresponding to the pixels within the marked area can be: ; in, It is the gray-level normalization factor corresponding to the i-th pixel in the marked area. i is the index of the pixel in the marked area. It is the grayscale value corresponding to the i-th pixel within the marked area. It is the maximum gray value of all pixels in all images of the image sequence to which the marked region belongs. It is the minimum gray value of all pixels in all images of the image sequence to which the marked region belongs.

[0027] For example, the formula for determining the gradient normalization factor corresponding to a pixel within a marked region can be: ; in, It is the gradient normalization factor corresponding to the i-th pixel in the marked region. i is the index of the pixel in the marked region. It is the gradient value corresponding to the i-th pixel within the marked area. It is the maximum gradient value corresponding to all pixels in all images of the image sequence to which the marked region belongs. It is the minimum gradient value corresponding to all pixels in all images of the image sequence to which the marked region belongs.

[0028] The second step is to determine the initial calcified patch feature factors corresponding to the above-mentioned marked area based on the mean and standard deviation of the gray-level normalization factors corresponding to all pixels in the above-mentioned marked area, and the mean of the gradient normalization factors corresponding to all edge pixels in the above-mentioned marked area.

[0029] For example, the formula for determining the initial calcified plaque feature factor corresponding to the marked region can be: ; Where E is the initial calcified patch feature factor corresponding to the marked region. T is the mean of the gradient normalization factor corresponding to all edge pixels of the marked region. It is an exponential function with the natural constant as its base. It is the mean of the gray-level normalization factors corresponding to all pixels within the marked area. It is the standard deviation of the gray-level normalization factor corresponding to all pixels in the marked area.

[0030] It should be noted that calcified plaques often contain calcium salt deposits, which typically have low water content and lack mobile protons. This results in lower signal intensity (lower grayscale) in the calcified portion on MRI sequences, and the grayscale distribution is often relatively stable. The signal intensity difference between the calcified plaque and adjacent tissues is often significant, leading to sharp and clear edges. Therefore, the edge gradient of calcified plaques is often large. A larger T value generally indicates a larger edge gradient in the marked area, and thus, a more likely sharp and clear edge. The smaller the value of E, the smaller the gray level of the marked area and the more stable the gray level distribution. Therefore, the larger the value of E, the greater the degree of calcification in the marked area, and the more likely the marked area is to represent calcified plaques.

[0031] The calcification stability analysis module 103 is used to perform calcification stability analysis based on the initial calcification plaque characteristic factors corresponding to different initial regions of carotid plaques, and to obtain the target calcification plaque characteristic factors corresponding to each initial region of carotid plaques.

[0032] As an example, determining the target calcified plaque characteristic factors corresponding to each initial region of a carotid plaque may include the following steps: The first step is to identify any initial region of a carotid plaque as a labeled region, and to identify the image sequence to which the labeled region belongs as a labeled image sequence. Then, to identify each image sequence in the target SWI image sequence, target T1 image sequence, and target T2 image sequence other than the labeled image sequence, two reference image sequences are obtained.

[0033] The second step is to select images from each reference image sequence whose image numbers are the same as the image numbers of the images to which the marked regions belong, and use these images as reference images, thus obtaining two reference images.

[0034] The third step, selecting the initial carotid plaque region that matches the marked region from each reference image, to obtain the matching region corresponding to each reference image, may include the following sub-steps: The first sub-step involves determining any reference image as the marker reference image, and selecting the region from the marker reference image that is in the same position as the marker region as the marker representative region.

[0035] The second sub-step involves selecting initial carotid plaque regions from the aforementioned marked reference images whose intersection with the regions represented by the aforementioned marks is not empty, as candidate regions.

[0036] In the third sub-step, if no candidate region exists, it is determined that the marked reference image does not have a corresponding matching region, which often indicates that the number of matching regions in the matching region set corresponding to the marked region is less than 2.

[0037] The fourth sub-step involves selecting the candidate region with the largest intersection area with the region represented by the above marker from all candidate regions, and using it as the matching region corresponding to the marker reference image.

[0038] The fourth step is to construct a set of matching regions corresponding to the marked regions from all the matching regions of the reference images.

[0039] Fifth step: If the number of matching regions in the above matching region set is less than 2, then set the target calcified plaque feature factor corresponding to the above marked region to the preset calcified plaque feature factor.

[0040] In practice, the appearance of true calcified areas in various MRI sequences is often relatively stable, while hemorrhagic areas, due to the paramagnetism of hemosiderin in the blood, often exhibit morphological abrupt changes on SWI, meaning there are usually sudden increases in the area's size and the degree of these increases is unstable. Furthermore, the longitudinal continuity of each channel is often spatially unstable, potentially leading to tomography. Therefore, calcified plaques tend to appear stably across multiple image sequences, while hemorrhagic plaques may exhibit tomography due to instability. When the number of matching regions in the matching region set is less than two, it often indicates that the carotid artery plaque represented by the labeled region has not appeared stably across multiple image sequences, and it is more likely to be a hemorrhagic plaque, with a lower probability of being calcified. The preset calcified plaque feature factor can be a pre-set threshold indicating a low probability of calcification, and it can be the minimum value within the target calcified plaque feature factor range. In this embodiment of the invention, the target calcified plaque feature factor can range from 0 to 1; therefore, in this embodiment, the preset calcified plaque feature factor can be 0.

[0041] Step 6: If the number of matching regions in the above matching region set is equal to 2, then the target calcified plaque feature factor corresponding to the above-mentioned marked region is determined based on the mean and range of the initial calcified plaque feature factors corresponding to the above-mentioned marked region and the two matching regions in the above matching region set.

[0042] When the number of matching regions in the matching region set is equal to 2, the three initial carotid plaque regions in the labeling region and the matching region set often represent the same carotid plaque, and the three initial carotid plaque regions correspond to different image types, namely SWI image, T1 image and T2 image respectively.

[0043] It should be noted that when the number of matching regions in the matching region set is equal to 2, it often indicates that the carotid artery plaque represented by the marked region is displayed in multiple image sequences. However, not only calcified plaques can be displayed in multiple image sequences, but hemorrhagic plaques, although they may have tomographic features, can also be displayed in multiple image sequences. Therefore, when the carotid artery plaque represented by the marked region is displayed in multiple image sequences, it is necessary to further determine, based on calcification features, whether the carotid artery plaque represented by the marked region is very likely to be a calcified plaque.

[0044] For example, if the number of matching regions in the matching region set is equal to 2, then the formula for determining the feature factor of the target calcified plaque corresponding to the marked region can be: ; in, It is the characteristic factor of the target calcified plaque corresponding to the marked region. It is the mean of the initial calcified plaque feature factors corresponding to the two matching regions in the labeled region and matching region sets, that is, the mean of the initial calcified plaque feature factors corresponding to the three initial carotid plaque regions in the labeled region and matching region sets. It is an exponential function with the natural constant as its base. It is the range of the initial calcified plaque feature factors corresponding to the two matching regions in the labeled region and the matching region set, that is, the range of the initial calcified plaque feature factors corresponding to the three initial carotid plaque regions in the labeled region and the matching region set. It is the maximum value among the initial calcified plaque feature factors corresponding to the three initial carotid plaque regions, which are the marked region and the matching region set. It is the minimum value among the initial calcified plaque feature factors corresponding to the three initial carotid plaque regions, which are the marked region and the matching region set.

[0045] It should be noted that the degree of calcification in calcified plaques is often high and relatively stable across different image sequences. When Larger values ​​often indicate a higher degree of calcification in the carotid artery plaque across different image sequences. The smaller the size, the more stable the calcification of the carotid artery plaque tends to be across different image sequences. Therefore, when The larger the value, the more likely the carotid plaque represented by the set of marked and matched regions is to be a calcified plaque.

[0046] The vulnerable hemorrhage plaque probability determination module 104 is used to determine the probability of vulnerable hemorrhage plaques corresponding to each carotid plaque initial region based on the distribution differences of the initial regions of carotid plaques in the target SWI image sequence, target T1 image sequence and target T2 image sequence, as well as the target calcified plaque characteristic factors corresponding to each initial region of carotid plaques, in order to assist in stroke risk stratification and early warning.

[0047] As an example, determining the likelihood of a vulnerable hemorrhage plaque corresponding to the initial region of each carotid artery plaque may include the following steps: The first step, identifying any initial region of a carotid artery plaque as a labeled region, and selecting a set of reference regions corresponding to the labeled region from the image sequence to which the labeled region belongs, may include the following sub-steps: The first sub-step involves determining the image sequence to which the marked region belongs as the marked image sequence, and determining the image to which the marked region belongs as the image to be determined.

[0048] The second sub-step involves determining the image preceding and following the image to be determined in the above-mentioned marked image sequence as temporary images, and selecting the region with the same position as the marked region from the above-mentioned temporary images as a temporary representative region.

[0049] It should be noted that if the image to be determined is an image located at the end of the labeled image sequence, then there is often only one temporary image. For example, if the image to be determined is the last image in the labeled image sequence, then there is often only one temporary image, the image preceding the image to be determined.

[0050] The third sub-step involves selecting initial carotid plaque regions from the temporary images whose intersection with the temporary representative regions is not empty, and using these regions as reference regions.

[0051] The fourth sub-step involves selecting the reference region with the largest intersection area with the aforementioned temporary representative region from all reference regions, and using it as the reference region corresponding to the aforementioned temporary image, thereby obtaining the reference regions corresponding to the previous image and the next image of the image to be determined.

[0052] The fifth sub-step involves constructing a set of reference regions corresponding to the marked regions by combining the reference regions corresponding to the preceding and following images of the image to be determined.

[0053] It should be noted that if the reference region set corresponding to the marked region contains two initial carotid plaque regions, it usually indicates that the carotid plaque represented by the marked region is displayed in both the preceding and following images of the image to be determined. If the reference region set corresponding to the marked region contains one initial carotid plaque region, it usually indicates that the carotid plaque represented by the marked region is displayed in an adjacent image of the image to be determined. If the reference region set corresponding to the marked region does not contain any initial carotid plaque regions, it usually indicates that the carotid plaque represented by the marked region has tomographic visualization, and it is very likely a hemorrhagic plaque.

[0054] The second step is to construct a set containing only a preset continuity index if the set of reference regions corresponding to the marked regions is empty. This set serves as the set of vertical spatial continuity indices corresponding to the marked regions.

[0055] It should be noted that when the reference region set corresponding to the marked region is empty, it often indicates that the carotid artery plaque represented by the marked region has tomographic visualization; the worse the continuity, the more likely it is a hemorrhagic plaque. The preset continuity index can be a pre-set threshold for what is considered poor continuity, and it can be the minimum value of the longitudinal spatial continuity index. In this embodiment of the invention, the value range of the longitudinal spatial continuity index can be 0-1; therefore, in this embodiment of the invention, the preset continuity index can be 0.

[0056] Third, if the set of reference regions corresponding to the marked regions is not empty, then based on the area overlap between the marked regions and each reference region in the set of reference regions, determine the vertical spatial continuity index corresponding to each reference region. Obtaining the set of vertical spatial continuity indices corresponding to the marked regions may include the following sub-steps: The first sub-step is to determine the intersection area of ​​the temporary representative area and each reference area as the target overlap area between the marked area and each reference area.

[0057] The second sub-step involves determining the longitudinal spatial continuity index corresponding to each reference region based on the target overlap area between the marked region and each reference region, and the difference between the initial calcified plaque feature factors corresponding to the marked region and each reference region.

[0058] For example, the formula for determining the longitudinal spatial continuity index corresponding to the reference region can be: ; in, It is the longitudinal spatial continuity index corresponding to the j-th reference region. j is the index of the reference region in the set of reference regions corresponding to the marked region. It is the target overlap area between the marked region and the j-th reference region. S is the area of ​​the marked region. It is an exponential function with the natural constant as its base. It is an absolute value function. E is the initial calcified plaque characteristic factor corresponding to the marked region. It is the initial calcified plaque characteristic factor corresponding to the j-th reference region.

[0059] It should be noted that when A larger value generally indicates a greater degree of overlap between the marked region and the j-th reference region. The smaller the value, the more similar the calcification features are between the marked region and the j-th reference region. Therefore, when A larger value often indicates that the carotid plaque represented by the marked area and the j-th reference area is more likely to be stably displayed in the longitudinal direction, and that the carotid plaque may be more continuous in the longitudinal space.

[0060] The third sub-step involves taking the longitudinal spatial continuity indices corresponding to all reference regions in the set of reference regions corresponding to the marked regions, and forming a set of longitudinal spatial continuity indices corresponding to the marked regions.

[0061] The fourth step, based on the differences in the distribution of the initial carotid plaque region in the target SWI image sequence, target T1 image sequence, and target T2 image sequence, and the set of longitudinal spatial continuity indices corresponding to the above-mentioned marked regions, determines the initial vulnerable bleeding factor corresponding to the above-mentioned marked regions, which may include the following sub-steps: The first sub-step involves selecting initial carotid plaque regions that characterize the carotid plaques to which the aforementioned labeled regions belong from all initial carotid plaque regions in the target SWI image sequence, target T1 image sequence, and target T2 image sequence, thus forming a set of labeled SWI regions, a set of labeled T1 regions, and a set of labeled T2 regions corresponding to the aforementioned labeled regions.

[0062] The labeled SWI region set can include all initial carotid plaque regions corresponding to the carotid plaques in the target SWI image sequence. The labeled T1 region set can include all initial carotid plaque regions corresponding to the carotid plaques in the target T1 image sequence. The labeled T2 region set can include all initial carotid plaque regions corresponding to the carotid plaques in the target T2 image sequence.

[0063] It should be noted that the labeled SWI region set, labeled T1 region set, and labeled T2 region set can be obtained through manual annotation. Optionally, the labeled SWI region set, labeled T1 region set, and labeled T2 region set can also be obtained by identifying the initial region of the same carotid plaque in different image sequences using the Intersection over Union (IoU) algorithm or the Dice coefficient algorithm, thereby obtaining the labeled SWI region set, labeled T1 region set, and labeled T2 region set.

[0064] The second sub-step involves forming a target region group by combining regions with the same image sequence number from the labeled SWI region set, labeled T1 region set, and labeled T2 region set.

[0065] The number of elements in different target region groups can characterize the differences in the distribution of the initial region of carotid plaques in different image sequences.

[0066] It should be noted that if carotid artery plaques show a tomographic distribution, the number of elements in the target region group is often less than 3, resulting in different numbers of elements in different target region groups.

[0067] The third sub-step, determining the local vulnerable bleeding factor for each target region group based on the area of ​​the region within each target region group and its corresponding set of longitudinal spatial continuity indices, may include the following steps: First, any target region group is designated as a temporary region group. If the temporary region group contains fewer than 3 elements, the local vulnerable bleeding factor corresponding to the temporary region group is set as the preset vulnerable bleeding factor.

[0068] It should be noted that when there are fewer than 3 elements in the temporary region group, it often indicates that the carotid artery plaque represented by the temporary region group is more likely to have tomographic visualization, and it is very likely to be a hemorrhagic plaque. In this case, the local vulnerable bleeding factor can be set to a larger value, which can be set manually. For example, the preset vulnerable bleeding factor can be set to 1.

[0069] Next, if there are 3 elements in the temporary region group, the T1 continuity difference corresponding to the temporary region group is determined based on the difference between the set of longitudinal spatial continuity indices corresponding to the SWI region and the set of longitudinal spatial continuity indices corresponding to the T1 region; the T2 continuity difference corresponding to the temporary region group is determined based on the difference between the set of longitudinal spatial continuity indices corresponding to the SWI region and the set of longitudinal spatial continuity indices corresponding to the T2 region; and the local vulnerable bleeding factor corresponding to the temporary region group is determined based on the difference between the area of ​​the SWI region and the areas of the T1 and T2 regions, as well as the T1 and T2 continuity differences corresponding to the temporary region group.

[0070] Specifically, the SWI region within the temporary region group can be the initial carotid plaque region belonging to the SWI image within the temporary region group. The T1 region within the temporary region group can be the initial carotid plaque region belonging to the T1 image within the temporary region group. The T2 region within the temporary region group can be the initial carotid plaque region belonging to the T2 image within the temporary region group.

[0071] For example, the formula for determining the T1 continuous difference corresponding to a temporary region group can be: ; in, It is the T1 continuous difference corresponding to the temporary region group. It is an absolute value function. D is the mean of all elements in the set of longitudinal spatial continuity indices corresponding to the SWI regions within the temporary region group. It is the mean of all elements in the set of longitudinal spatial continuity indexes corresponding to region T1 within the temporary region group.

[0072] It should be noted that when A larger value often indicates a greater difference in the longitudinal spatial continuity index of carotid plaques between SWI and T1.

[0073] For example, the formula for determining the T2 continuous difference corresponding to the temporary region group can be: ; in, It is the T2 continuous difference corresponding to the temporary region group. It is an absolute value function. D is the mean of all elements in the set of longitudinal spatial continuity indices corresponding to the SWI regions within the temporary region group. It is the mean of all elements in the set of longitudinal spatial continuity indices corresponding to region T2 within the temporary region group.

[0074] It should be noted that when A larger value often indicates a greater difference in the longitudinal spatial continuity index of carotid plaques between SWI and T2.

[0075] For example, if there are 3 elements in a temporary region group, the formula for determining the corresponding local vulnerable bleeding factor for the temporary region group can be: ; in, It is the local vulnerable bleeding factor corresponding to the temporary regional group. This is the normalization function, mainly used for normalization. Depending on the specific situation, normalization can also be achieved using other normalization functions. L is the area of ​​the SWI region within the temporary region group. It is the average of the areas of region T1 and region T2 within the temporary region group. It is the T1 continuous difference corresponding to the temporary region group. It is the T2 continuous difference corresponding to the temporary region group.

[0076] In practice, vulnerable hemorrhagic plaques often exhibit significant expansion in the SWI channel, and their longitudinal spatial continuity in this sequence often differs considerably from that in T1 / T2 sequences. When A larger SWI value generally indicates that the area of ​​the carotid plaque represented by the temporary region group is more likely to be larger than the area of ​​the T1 / T2 channel, and that the carotid plaque represented by the temporary region group is more likely to be significantly dilated in the SWI channel. A larger value often indicates a greater difference in the longitudinal spatial continuity of the carotid plaque represented by the temporary region group compared to the spatial continuity of the T1 / T2 sequences. Therefore, when The larger the value, the more likely the carotid plaque characterized by the temporary regional group is to be a vulnerable hemorrhagic plaque.

[0077] The fourth sub-step involves determining the initial vulnerable bleeding factor corresponding to the carotid plaque to which the marked region belongs, based on the mean value of the local vulnerable bleeding factors corresponding to all target region groups and the number of target region groups with an element count of 3. The initial vulnerable bleeding factor corresponding to the carotid plaque to which the marked region belongs is then denoted as the initial vulnerable bleeding factor corresponding to the marked region.

[0078] For example, the formula for determining the initial vulnerable bleeding factor corresponding to the carotid artery plaque to which the marked area belongs can be: ; Wherein, F is the initial vulnerable bleeding factor corresponding to the carotid artery plaque to which the marked area belongs, that is, the initial vulnerable bleeding factor corresponding to the marked area. It is the mean of the local vulnerable bleeding factors corresponding to all target area groups. It is an exponential function with the natural constant as its base. n is the number of target region groups with 3 elements. N is the total number of all target region groups.

[0079] It should be noted that when A larger value generally indicates that the carotid artery plaque represented by the marked area is more likely to be a vulnerable hemorrhagic plaque. The smaller the value of F, the more likely the carotid artery plaque represented by the marked area is to have tomographic visualization in multiple image sequences. Therefore, the larger the value of F, the more likely the carotid artery plaque represented by the marked area is to be a vulnerable hemorrhagic plaque.

[0080] The fifth step is to determine the likelihood of a vulnerable hemorrhage plaque corresponding to the marked area based on the initial vulnerable hemorrhage factor and the target calcified plaque characteristic factor corresponding to the marked area.

[0081] For example, the formula for determining the probability of a vulnerable hemorrhage plaque corresponding to a marked area can be: ; in, This represents the probability of a vulnerable hemorrhage plaque corresponding to the marked area. F represents the initial vulnerable hemorrhage factor corresponding to the carotid artery plaque to which the marked area belongs, i.e., the initial vulnerable hemorrhage factor corresponding to the marked area. It is an exponential function with the natural constant as its base. It is the characteristic factor of the target calcified plaque corresponding to the marked region.

[0082] It should be noted that a larger F value generally indicates that the carotid artery plaque represented by the marked area is more likely to be a vulnerable hemorrhagic plaque. The larger the value, the more likely the carotid artery plaque represented by the set of marked and matched regions is to be a calcified plaque rather than a vulnerable hemorrhagic plaque. Therefore, when The larger the value, the more likely the carotid artery plaque represented by the marked area is to be a vulnerable hemorrhage plaque. To a certain extent, this can help screen for vulnerable hemorrhage plaques in the carotid artery, rather than mistaking them for calcified plaques.

[0083] Optionally, embodiments of the present invention may further include the following steps: The first step is to obtain all the initial regions of the carotid artery plaque in the target SWI image sequence, all the initial regions of the carotid artery plaque in the target T1 image sequence, and all the initial regions of the carotid artery plaque in the target T2 image sequence, thus forming a set of initial regions of the carotid artery plaque corresponding to the carotid artery plaque.

[0084] The second step is to determine the overall vulnerable bleeding probability for each carotid plaque by taking the mean of the vulnerability bleeding plaque probability corresponding to all initial regions of carotid plaques in the initial region set corresponding to each carotid plaque.

[0085] The third step is to determine if the overall vulnerable bleeding probability corresponding to the carotid artery plaque is greater than the preset bleeding probability threshold. In this case, a higher stroke risk can often be set as an early warning.

[0086] The preset bleeding probability threshold can be the maximum allowable comprehensive vulnerable bleeding probability when a carotid artery plaque is considered a calcified plaque, set in advance based on actual conditions; for example, it can be 0.5. The preset bleeding probability threshold can be obtained by acquiring the comprehensive vulnerable bleeding probability corresponding to multiple historical calcified plaques, and determining the maximum value of the comprehensive vulnerable bleeding probability corresponding to all historical calcified plaques as the preset bleeding probability threshold. Historical calcified plaques can be carotid artery plaques that have been confirmed as calcified plaques. In practice, the more different historical calcified plaques there are, the more reasonable the preset bleeding probability threshold setting tends to be.

[0087] The fourth step is to determine if the overall vulnerable bleeding probability corresponding to the carotid artery plaque is less than or equal to the preset bleeding probability threshold. In this case, a smaller stroke risk can often be set as a warning.

[0088] Optionally, if the patient undergoes multiple tests, the results can be used to determine whether the patient's condition has worsened. These test results may include: the overall vulnerable bleeding probability corresponding to the carotid artery plaque, the maximum stenosis index of all carotid artery plaques in the patient, and the degree of deviation of monitored indicators from normal. Monitored indicators may include, but are not limited to, triglycerides and fasting blood glucose. Specifically, if the overall vulnerable bleeding probability gradually increases with the number of tests, the maximum stenosis index of all carotid artery plaques in the patient gradually increases, or the degree of deviation of monitored indicators from normal gradually increases, it often indicates that the patient's stroke risk is gradually increasing.

[0089] refer to Figure 2 Based on the same inventive concept as the above-described method embodiments, this invention provides a method for intelligent stratification and early warning of stroke risk based on carotid plaque characteristics, comprising the following steps: Step S1: Obtain the target SWI image sequence, target T1 image sequence, and target T2 image sequence corresponding to the patient to be tested after multi-sequence registration, and extract the region of each image in these image sequences to obtain the initial region of carotid plaque.

[0090] Step S2: Analyze the gray values ​​and gradient values ​​of the pixels in the initial region of each carotid plaque to determine the initial calcified plaque feature factors corresponding to the initial region of each carotid plaque.

[0091] Step S3: Calcification stability analysis is performed based on the initial calcified plaque characteristic factors corresponding to different initial regions of carotid plaques to obtain the target calcified plaque characteristic factors corresponding to each initial region of carotid plaque.

[0092] Step S4: Based on the differences in the distribution of the initial regions of carotid plaques in the target SWI image sequence, target T1 image sequence, and target T2 image sequence, and the target calcified plaque characteristic factors corresponding to each initial region of carotid plaque, determine the likelihood of vulnerable hemorrhage plaques corresponding to each initial region of carotid plaque, which is used to assist in stroke risk stratification and early warning.

[0093] Figure 3 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. For example, as shown... Figure 3 As shown, the computer device 300 includes: a memory 301, a processor 302, and a computer program 303 stored in the memory 301 and running on the processor 302. When the processor 302 executes the computer program 303, the computer device can execute the aforementioned intelligent stratification and early warning method for stroke risk based on carotid plaque characteristics.

[0094] In summary, this invention quantifies the initial calcified plaque characteristic factors corresponding to the initial region of the carotid plaque by analyzing the grayscale and gradient of the initial region of the carotid plaque, and performs calcification stability analysis to quantify the target calcified plaque characteristic factors corresponding to the initial region of the carotid plaque. Based on the distribution differences of the initial region of the carotid plaque in the target SWI image sequence, target T1 image sequence, and target T2 image sequence, and the target calcified plaque characteristic factors corresponding to the initial region of the carotid plaque, the probability of the initial region of the carotid plaque being a vulnerable hemorrhagic plaque is quantified. The larger the value, the more likely the initial region of the carotid plaque is to represent a vulnerable hemorrhagic plaque. To a certain extent, this can accurately assist in the identification of calcified plaques and vulnerable hemorrhagic plaques, improve the accuracy of identification, and thus improve the auxiliary effect of stroke risk warning.

Claims

1. A stroke risk intelligent stratification early warning system based on carotid plaque characteristics, characterized in that, The system includes: The image acquisition and region extraction module is used to acquire the target SWI image sequence, target T1 image sequence and target T2 image sequence corresponding to the patient to be tested after multi-sequence registration, and to extract the region of each image in these image sequences to obtain the initial region of carotid plaque; The grayscale and gradient analysis module is used to analyze the grayscale and gradient values ​​of the pixels in the initial region of each carotid plaque to determine the initial calcified plaque feature factors corresponding to the initial region of each carotid plaque. The calcification stability analysis module is used to perform calcification stability analysis based on the initial calcification plaque characteristic factors corresponding to different initial regions of carotid plaques, and to obtain the target calcification plaque characteristic factors corresponding to each initial region of carotid plaque. The vulnerable hemorrhage plaque probability determination module is used to determine the probability of vulnerable hemorrhage plaques corresponding to each initial region of carotid artery plaque based on the distribution differences of the initial regions of carotid artery plaques in the target SWI image sequence, target T1 image sequence, and target T2 image sequence, as well as the target calcified plaque feature factors corresponding to each initial region of carotid artery plaque, in order to assist in stroke risk stratification and early warning.

2. The intelligent stratified early warning system for stroke risk based on carotid plaque characteristics according to claim 1, characterized in that, The process of extracting regions from each of these image sequences to obtain the initial region of the carotid plaque includes: Threshold segmentation is performed on each image in the target SWI image sequence, target T1 image sequence, and target T2 image sequence to obtain the initial region of the carotid plaque that characterizes the carotid plaque.

3. The intelligent stratified early warning system for stroke risk based on carotid plaque characteristics according to claim 1, characterized in that, The analysis, based on the grayscale and gradient values ​​of pixels within the initial region of each carotid plaque, determines the initial calcified plaque feature factors corresponding to each initial region of the carotid plaque, including: An initial region of any carotid artery plaque is defined as a labeled region, and the gray value and gradient value corresponding to each pixel in the labeled region are normalized to obtain the gray value normalization factor and gradient normalization factor corresponding to each pixel in the labeled region. The initial calcification patch feature factor corresponding to the marked region is determined based on the mean and standard deviation of the gray-level normalization factor corresponding to all pixels in the marked region, and the mean of the gradient normalization factor corresponding to all edge pixels in the marked region.

4. The intelligent stratified early warning system for stroke risk based on carotid plaque characteristics according to claim 1, characterized in that, The step involves analyzing calcification stability based on initial calcified plaque characteristic factors corresponding to different initial carotid plaque regions to obtain target calcified plaque characteristic factors for each initial carotid plaque region, including: Any initial region of a carotid artery plaque is identified as a labeled region, and the image sequence to which the labeled region belongs is identified as a labeled image sequence. Each image sequence in the target SWI image sequence, target T1 image sequence, and target T2 image sequence, except for the labeled image sequence, is identified as a reference image sequence, resulting in two reference image sequences. From each reference image sequence, images with the same image number as the image to which the marked region belongs are selected as reference images, resulting in two reference images; Initial carotid plaque regions that match the marked regions are selected from each reference image to obtain the matching regions corresponding to each reference image; The matching regions corresponding to all reference images are used to form a set of matching regions corresponding to the marked region; If the number of matching regions in the matching region set is less than 2, then the target calcified plaque feature factor corresponding to the marked region is set to the preset calcified plaque feature factor. If the number of matching regions in the matching region set is equal to 2, then the target calcified plaque feature factor corresponding to the marked region is determined based on the mean and range of the initial calcified plaque feature factors corresponding to the marked region and the two matching regions in the matching region set.

5. The intelligent stratified early warning system for stroke risk based on carotid plaque characteristics according to claim 4, characterized in that, The step of selecting initial carotid plaque regions that match the marked region from each reference image to obtain a matching region for each reference image includes: Any reference image is selected as the marker reference image, and the region with the same position as the marker region is selected from the marker reference image as the marker representative region; Initial regions of carotid plaques that do not intersect with the marked regions in the labeled reference image are selected as candidate regions. If no candidate region exists, it is determined that the labeled reference image does not have a corresponding matching region; If a candidate region exists, the candidate region with the largest intersection area with the region represented by the marker is selected from all candidate regions and used as the matching region corresponding to the marker reference image.

6. The intelligent stratified early warning system for stroke risk based on carotid plaque characteristics according to claim 1, characterized in that, The determination of the likelihood of a vulnerable hemorrhagic plaque corresponding to each initial carotid plaque region based on the distribution differences of the initial carotid plaque regions in the target SWI image sequence, target T1 image sequence, and target T2 image sequence, and the target calcified plaque characteristic factors corresponding to each initial carotid plaque region, includes: An initial region of any carotid artery plaque is identified as a marker region, and a set of reference regions corresponding to the marker region is selected from the image sequence to which the marker region belongs. If the set of reference regions corresponding to the marked region is empty, then a set containing only a preset continuity index is constructed as the set of vertical spatial continuity indices corresponding to the marked region. If the set of reference regions corresponding to the marked region is not empty, then the longitudinal spatial continuity index corresponding to each reference region is determined according to the area overlap between the marked region and each reference region in the set of reference regions, and the set of longitudinal spatial continuity indices corresponding to the marked region is obtained. Based on the differences in the distribution of the initial region of carotid plaque in the target SWI image sequence, target T1 image sequence, and target T2 image sequence, and the set of longitudinal spatial continuity indices corresponding to the marked region, the initial vulnerable bleeding factor corresponding to the marked region is determined. The likelihood of a vulnerable hemorrhage plaque corresponding to the marked region is determined based on the initial vulnerable hemorrhage factor and the target calcified plaque characteristic factor corresponding to the marked region.

7. The intelligent stratified early warning system for stroke risk based on carotid plaque characteristics according to claim 6, characterized in that, The step of filtering the reference region set corresponding to the marked region from the image sequence to which the marked region belongs includes: The image sequence to which the marked region belongs is determined as the marked image sequence, and the image to which the marked region belongs is determined as the undetermined image; The image preceding and following the image to be determined in the marked image sequence are identified as temporary images, and the region with the same position as the marked region is selected from the temporary images as a temporary representative region. Initial carotid plaque regions with non-empty intersections with the temporary representative region are selected from the temporary image and used as reference regions; The reference region with the largest intersection area with the temporary representative region is selected from all reference regions and used as the reference region corresponding to the temporary image, thereby obtaining the reference regions corresponding to the previous image and the next image of the image to be determined. The reference regions corresponding to the preceding and following images of the image to be determined are used to form a set of reference regions corresponding to the marked region.

8. The intelligent stratified early warning system for stroke risk based on carotid plaque characteristics according to claim 7, characterized in that, The step of determining the vertical spatial continuity index corresponding to each reference region based on the area overlap between the marked region and each reference region in the corresponding set of reference regions includes: The area of ​​the intersection between the temporary representative area and each reference area is determined as the target overlap area between the marked area and each reference area; Based on the target overlap area between the marked region and each reference region, and the difference between the initial calcified plaque characteristic factors corresponding to the marked region and each reference region, the longitudinal spatial continuity index corresponding to each reference region is determined.

9. A stroke risk intelligent stratification early warning system based on carotid plaque characteristics according to claim 6, characterized in that, The step of determining the initial vulnerable bleeding factor corresponding to the marked region based on the distribution differences of the initial carotid plaque region in the target SWI image sequence, target T1 image sequence, and target T2 image sequence, and the set of longitudinal spatial continuity indices corresponding to the marked region, includes: From all initial regions of carotid plaques in the target SWI image sequence, target T1 image sequence and target T2 image sequence, the initial regions of carotid plaques that characterize the carotid plaques to which the labeled region belongs are selected to form the labeled SWI region set, labeled T1 region set and labeled T2 region set corresponding to the labeled region; Regions with the same image sequence number in the labeled SWI region set, labeled T1 region set, and labeled T2 region set are formed into target region groups. The number of elements in different target region groups represents the differences in the distribution of the initial region of carotid plaque in different image sequences. Based on the area of ​​each target region group and its corresponding set of longitudinal spatial continuity indices, the local vulnerable bleeding factors corresponding to each target region group are determined. Based on the mean value of the local vulnerable bleeding factors corresponding to all target region groups and the number of target region groups with an element count of 3, the initial vulnerable bleeding factor corresponding to the carotid plaque to which the marked region belongs is determined. The initial vulnerable bleeding factor corresponding to the carotid artery plaque to which the marked region belongs is denoted as the initial vulnerable bleeding factor corresponding to the marked region.

10. A stroke risk intelligent stratification early warning system based on carotid plaque characteristics according to claim 9, characterized in that, The determination of local vulnerable bleeding factors for each target region group based on the area of ​​the region within each target region group and its corresponding set of longitudinal spatial continuity indices includes: Any target region group is designated as a temporary region group. If there are fewer than 3 elements in the temporary region group, the local vulnerable bleeding factor corresponding to the temporary region group is set as a preset vulnerable bleeding factor. If there are 3 elements in the temporary region group, the T1 continuity difference corresponding to the temporary region group is determined based on the difference between the set of longitudinal spatial continuity indices corresponding to the SWI region and the set of longitudinal spatial continuity indices corresponding to the T1 region; the T2 continuity difference corresponding to the temporary region group is determined based on the difference between the set of longitudinal spatial continuity indices corresponding to the SWI region and the set of longitudinal spatial continuity indices corresponding to the T2 region; and the local vulnerable bleeding factor corresponding to the temporary region group is determined based on the difference between the area of ​​the SWI region and the areas of the T1 and T2 regions, as well as the T1 and T2 continuity differences corresponding to the temporary region group.

Images